Play
Search post

The Curse of Dead Lambda Functions

0
Most developers agree that AWS Lambda is a great service. It’s cheap and easy to start with. It’s scalable. And most importantly — it just works. 

But there are challenges as well. One the common challenges is to keep track of all your functions, which often leads to “dead” functions. 

A dead Lambda function is a function that has been cold for a long time (makes sense, doesn’t it?). Usually this function is useless, and should probably be removed. 

But it’s not just about keeping your resources nice and clean — there are real issues with dead functions:
  1. Management: code and error management are common issues in every complex serverless architecture — but when you add dozens of unused functions, things become much more complicated.
  2. Security: more functions means more opportunities for an attacker to try and take advantage of your code. A function that hasn’t been used in a long time is most likely outdated and may have wrong permissions or contain old external libraries.
  3. Performance analysis: when trying to obtain a deep understanding of the entire architecture, such as the end-to-end performance, or the costs of specific transactions, having unused resources makes everything a lot harder.

So we can agree that dead Lambdas are bad. What can we do about it? 


Dead AWS Lambda thoughts

Solution

In order to solve this problem, we at Epsagon have created an open-source project: list-lambdas. It’s a Python script which enumerates your Lambda functions across all AWS regions and provides useful information, such as the function’s memory settings, its last invocation time and more. It can help identifying dead Lambda functions easily.

Setup

git clone git@github.com:epsagon/list-lambdas.git 
cd list-lambdas/ 
pip install -r requirements.txt 
python list_lambdas.py

Example Outputs

CLI output

CSV output

Example Usage

The usage is straightforward using the CLI. 

Filter only Lambda functions that have not been active in the last 10 days: 
python list_lambdas.py --inactive-days-filter 10 

Output table (with extra data) to a CSV file: 
python list_lambdas.py --csv lambdas.csv 

Print extended information to the screen (same as in the CSV file): 
python list_lambdas.py --all 

Sort by a chosen column (e.g. by last invocation time): 
python list_lambdas.py --sort-by last-invocation

Conclusion

Dead serverless functions are a real issue — but it can be handled rather easily. Best practices such as giving your functions meaningful names are relevant in serverless as well. Tools like list-lambdas and others can help keep your architecture clean and steady, which ultimately leads to better software. We will keep adding features to the project, and plan to release additional projects in the near future. 

We are curious for your feedback! Do you have additional features in mind? Please tell us about them, and feel free to contribute to the project!
How to Make Lambda Faster: Memory Performance Benchmark

About Author

Nitzan Shapira
Nitzan Shapira

I am a software engineer with over 12 years of experience in programming, machine learning, cyber-security, and reverse engineering. During the last year, I am working on Epsagon, being a co-founder and the CEO. Epsagon is focused on bringing observability to serverless cloud applications, using distributed tracing and AI technologies, tackling the unique challenges in such environments. Apart from co-founding Epsagon, I enjoy playing the piano, I am a traveling enthusiast, an experienced chess player, and addicted to sports.

Related Posts
How to Package External Code in AWS Lambda Using Serverless Framework
Stackery and Epsagon to Increase Serverless Observability
More than Functions - Serverless Observability Webinar

Comment

Subscribe To Blog

Subscribe to Email Updates